Previously we couldn't use == to compare ulogs, so we had to do __dict__
comparison manually in tests. We should just add such comparison as
default, since two ULog objects that have the same data should be
considered equal.

Writing DatabaseULog to file doesn't work, so we add a test which finds
errors we can fix.

If DatabaseULog is lazy loaded it doesn't have data, so it will fail if
trying to write to file.

DatabaseULog.write_ulog complained about wrong value types during struct
packing.

There were lost in pyulog.2.sql, so the DB is very slow for large
databases.

NB: Applying this migration will take a while for large databases.